Skip to main content
ARS Home » Research » Publications at this Location » Publication #241216
Title: Stereo Spectral Imaging System for Plant Health Characterization

Author
item Yoon, Seung-Chul
item THAI, CHI - University Of Georgia

Submitted to: ASABE Annual International Meeting
Publication Type: Proceedings
Publication Acceptance Date: 11/13/2008
Publication Date: 6/21/2009
Citation: Yoon, S.C., Thai, C. 2009. Stereo Spectral Imaging System for Plant Health Characterization. ASABE Annual International Meeting. Technical Paper No. 096583.

Interpretive Summary: There is a growing need for developing 3D imaging techniques applied to characterization of plant health. Conventional 3D imaging involves measurements of distance information of a surface point via either stereo vision or a range finder and also involves 3D surface modeling and reconstruction. Although conventional 3D imaging of plants is useful to study disease propagation paths or any observable structure characteristics of plants, 3D spatial information alone is not enough to determine plant health. As supplementary but essential technology for plant health characterization, spectral imaging can measure spectral information directly related to plant health, such as chlorophyll concentration in leaves and plant productivity. In this study, a 3D spectral imaging system was developed to explore the possibility of measuring spatial and spectral information at the same time, to process them, and to visualize processed data in 3D space. The developed imaging system (called a stereo spectral imaging system) consisted of a stereo vision camera to obtain 3D images and two bandpass optical filters to acquire spectral reflectance images and custom software to process and to visualize the data. The bandpass filters were attached to the stereo vision camera to capture stereo spectral images at two different wavelengths, 690 nm (red) and 750 nm (near-infrared). The spectral images were processed to produce the normalized difference vegetation index (NDVI) values that were calculated from calibrated reflectance values. The software produced virtual 3D computer graphics images in which modeled plant surfaces were rendered in NDVI values. The study with an artificial organic plant sample demonstrated the potential of the stereo spectral imaging system to monitor plant health in real time 3D. Further research is needed to improve the accuracy of 3D plant modeling and to evaluate the imaging system with real plants in both indoor and outdoor environments.

Technical Abstract: Three-dimensional (3D) measurements of whole plants may provide detailed structure information about plant health and also complement existing X-ray systems for the below-ground part. In addition to this structure characterization of plants, spectral information may also biochemically characterize plants' health. A stereo vision technique is a cost-effective and rapid imaging technique for measuring and reconstructing 3D structures. The Normalized Difference Vegetation Index (NDVI) requiring measurements of two spectral wavelengths at NIR and red spectral regions has been widely used in remote sensing as an index to estimate various vegetation properties including chlorophyll concentration in leaves, leaf area index, biomass, and plant productivity. We integrated both stereo vision and NDVI techniques and developed a stereo spectral imaging (SSI) system for chlorophyll and biomass quantification of plants in 3D space. We used a stereo vision camera system and custom designed a dual-wavelength filter system at 690 nm and 750 nm to develop the SSI system. Calibration techniques for NDVI computation using two spectral band images were developed by referencing a diffuse reflectance panel. We also developed a texture mapping technique for rendering NDVI values in 3D space. In this paper, the performance of the SSI system was evaluated with an artificial plant showing spectral properties similar to green leaves of real plants.